Title:Soft Interference Cancellation for Random Coding in Massive Gaussian Multiple-Access

Abstract:We utilize recent results on the exact block error probability of Gaussian random codes in additive white Gaussian noise to analyze Gaussian random coding for massive multiple-access at finite message length. Soft iterative interference cancellation is found to closely approach the performance bounds recently found in [1]. The existence of two fundamentally different regimes in the trade-off between power and bandwidth efficiency reported in [2] is related to much older results in [3] on power optimization by linear programming. Furthermore, we tighten the achievability bounds of [1] in the low power regime and show that orthogonal constellations are very close to the theoretical limits for message lengths around 100 and above.